cFineGAN: Unsupervised multi-conditional fine-grained image generation

TL;DR

cFineGAN is an unsupervised image generation method that combines texture and shape from two images using shape-biased models, demonstrated on multiple datasets.

Contribution

It extends FineGAN to multi-conditional generation with shape-biased models, enabling texture-shape controlled image synthesis without supervision.

Findings

Shape-biased models improve generation quality.

Effective across multiple benchmark datasets.

Unsupervised approach achieves promising results.

Abstract

We propose an unsupervised multi-conditional image generation pipeline: cFineGAN, that can generate an image conditioned on two input images such that the generated image preserves the texture of one and the shape of the other input. To achieve this goal, we extend upon the recently proposed work of FineGAN \citep{singh2018finegan} and make use of standard as well as shape-biased pre-trained ImageNet models. We demonstrate both qualitatively as well as quantitatively the benefit of using the shape-biased network. We present our image generation result across three benchmark datasets- CUB-200-2011, Stanford Dogs and UT Zappos50k.